Legal claims defining the scope of protection. Each claim is shown in both the original legal language and a plain English translation.
1. An image processing apparatus, comprising: a processor configured to determine whether an input image is a low-quality image or a high-quality image; determine a first number of clearest images from among a plurality of low-quality images being determined; and perform a face detection on a determined clearest image and a determined high-quality image using a second convolutional neural network (CNN) classifier model, and output a result image resulting from the performing of the face detection, wherein the result image is a face image or an image in which a face is absent.
Image processing for quality assessment and face detection. This invention addresses the challenge of accurately detecting faces in images, particularly when dealing with varying image quality. The apparatus includes a processor that first classifies an input image as either low-quality or high-quality. If the image is determined to be low-quality, the processor then identifies a specific number of the clearest images from a set of multiple low-quality images. Subsequently, face detection is performed on either the determined clearest low-quality image or the determined high-quality image. This face detection is carried out using a second convolutional neural network (CNN) classifier model. The output of this process is a result image, which will either contain a detected face or indicate the absence of a face.
2. The image processing apparatus of claim 1 , wherein the processor is further configured to calculate an image classification probability value used to classify the input image as a type among a plurality of different quality types using a first CNN classifier model, and determine whether the input image is the low-quality image or the high-quality image based on the image classification probability value of the input image.
3. The image processing apparatus of claim 1 , wherein the processor is further configured to calculate an image clearness with respect to each of the low-quality images and to determine a first number of low-quality images from among the low-quality images in an order from a highest clearness, to be the first number of clearest images.
This invention relates to image processing systems designed to enhance the quality of low-quality images, particularly in scenarios where multiple low-quality images of the same scene are available. The problem addressed is the challenge of selecting the clearest images from a set of low-quality images to improve overall image quality, such as in surveillance, medical imaging, or low-light photography. The apparatus includes a processor that processes multiple low-quality images of the same scene. The processor calculates an image clearness metric for each low-quality image, which quantifies the sharpness, contrast, or other quality indicators of the image. Based on these clearness values, the processor selects a predefined number of the clearest images from the set. This selection process prioritizes images with the highest clearness, ensuring that the most visually distinct or high-quality images are chosen for further processing or analysis. The selected images can then be used individually or combined to produce a higher-quality output, such as through super-resolution techniques or image fusion. The invention improves upon existing methods by providing an automated and objective way to identify the best-quality images from a set, reducing manual selection and improving efficiency in applications where image clarity is critical.
4. The image processing apparatus of claim 1 , the processor is further configured to: determine whether a face image, of plural face images output by the face detector, is a blurred face image; and determine a second predetermined number of clearest face images among a plurality of blurred face images being determined.
5. The image processing apparatus of claim 4 , wherein the processor is configured to calculate a face image classification probability value used to classify the face image as a clear face image or the blurred face image using a third CNN classifier model, and determine whether the face image is the clear face image or the blurred face image based on the face image classification probability value.
6. The image processing apparatus of claim 4 , wherein the processor is configured to calculate a face clearness with respect to each of the plurality of blurred face images, and determine a second predetermined number of blurred face images among the plurality of blurred face images in an order from a highest face clearness, to be the second predetermined number of clearest images.
This invention relates to image processing systems designed to enhance facial clarity in images, particularly when multiple blurred face images are present. The problem addressed is the difficulty in automatically selecting the clearest face images from a set of blurred or low-quality images, which is critical for applications like facial recognition, surveillance, or biometric analysis. The system includes a processor that evaluates a collection of blurred face images. For each image, the processor calculates a "face clearness" metric, which quantifies the sharpness or visibility of the facial features. The processor then ranks the images based on this metric and selects a predetermined number of the clearest images from the ranked set. This selection process ensures that only the highest-quality face images are retained for further processing, improving accuracy in subsequent applications like identification or analysis. The invention builds on a prior system that generates multiple blurred face images by applying different blur levels to an original image. The processor in this enhanced system further refines the selection by prioritizing images with the highest face clearness, ensuring optimal clarity for downstream tasks. This approach is particularly useful in scenarios where multiple degraded versions of a face image exist, such as in low-light conditions or when images are captured from varying distances.
7. The method of claim 1 , wherein the processor is further configured to determine an image clearness with respect to each of the low-quality images based on an inverse of a standard deviation of the image classification probability value of the respective image.
8. An image processing apparatus comprising: a processor configured to: determine whether a face image is a blurred face image; and determine a second predetermined number of clearest face images among a plurality of blurred face images being determined.
9. The image processing apparatus of claim 8 , wherein the processor is configured to calculate a face image classification probability value used to classify the face image as one of a clear face image and the blurred face image using a convolutional neural network (CNN) classifier model, and determine whether the face image is the clear face image or the blurred face image based on the face image classification probability value of the face image.
10. An image processing method comprising: actuating a processor to: determine whether an input image is a low-quality image or a high-quality image; determine a first number of clearest images from determined low-quality images; and perform a face detection on a determined clearest image and a determined high-quality image using a first convolutional neural network (CNN) classifier model and outputting a result image resulting from the performing of the face detection, wherein the result image is a face image or an image in which a face is absent.
11. The image processing method of claim 10 , wherein the actuating of the processor to determine whether the input image is the low-quality image or the high-quality image comprises the actuating of the processor to: calculate an image classification probability value to classify the input image as a type among a plurality of different quality types using a second convolutional neural network (CNN) classifier model; and determine whether the input image is the low-quality image or the high-quality image based on the image classification probability value of the input image.
12. The image processing method of claim 11 , wherein the input image is captured by a camera operably coupled to the processor.
This invention relates to image processing systems that enhance image quality, particularly for images captured by cameras. The problem addressed is the need for efficient and accurate image processing techniques to improve visual clarity, especially in low-light or noisy environments. The method involves capturing an input image using a camera connected to a processor, followed by processing the image to reduce noise, correct distortions, or enhance features. The processor applies algorithms to analyze and modify the image data, such as adjusting brightness, contrast, or sharpness, and may also perform object detection or recognition. The processed image is then output for display or further analysis. The system ensures real-time or near-real-time processing, making it suitable for applications like surveillance, medical imaging, or autonomous vehicles. The invention focuses on optimizing image quality while minimizing computational overhead, ensuring compatibility with various camera types and processing hardware. The method may also include adaptive adjustments based on environmental conditions or user preferences to further enhance performance.
13. The image processing method of claim 10 , wherein the actuating of the processor to determine the first predetermined number of clearest images comprises the actuating of the processor to: calculate an image clearness with respect to each of the low-quality images; and determine a first number of low-quality images from among the low-quality images in an order from a highest clearness, to be the first predetermined number of clearest images.
14. The image processing method of claim 10 , further comprising the actuating of the processor to: determine whether the face image is a blurred face image; and determine a second predetermined number of clearest face images among a plurality of determined blurred face images.
15. The image processing method of claim 14 , wherein the actuating of the processor to determine whether the face image is the blurred face image comprises the actuating of the processor to: calculate a face image classification probability value used to classify the face image as a clear face image or the blurred face image using a third CNN classifier model; and determine whether the face image is the clear face image or the blurred face image based on the face image classification probability value.
16. The image processing method of claim 14 , wherein the actuating of the processor to determine the second predetermined number of clearest face images comprises the actuating of the processor to: calculate a face clearness with respect to each of the plurality of blurred face images; and determine a second predetermined number of blurred face images among the plurality of blurred face images in an order from a highest face clearness, to be the second predetermined number of clearest images.
17. A non-transitory computer readable storage medium storing instructions that, when actuated by a processor, cause the processor to perform the method of claim 10 .
18. An image processing method, comprising: actuating a camera to capture input images; actuating a processor to: selectively generate a subset of candidate images from the input images according to image quality; adaptively allocate processing resources to identify face images amongst the selectively generated subset of candidate images, wherein the identifying face images includes performing a face detection on the selectively generated subset of candidate images using a first convolutional neural network (CNN) classifier model and outputting a result image resulting from the performing of the face detection, wherein the result image is a face image or an image in which a face is absent.
This invention relates to image processing methods for efficiently detecting faces in captured images. The problem addressed is the computational inefficiency of processing all captured images for face detection, especially in resource-constrained environments. The method involves capturing input images using a camera and then selectively generating a subset of candidate images based on image quality metrics. This reduces the number of images that require further processing. A processor then adaptively allocates processing resources to identify face images within this subset. The face detection is performed using a first convolutional neural network (CNN) classifier model, which outputs a result image indicating either the presence of a face or the absence of a face. The adaptive resource allocation ensures that higher-quality images receive more processing attention, optimizing computational efficiency. The method is particularly useful in applications requiring real-time face detection, such as surveillance systems or biometric authentication, where processing speed and accuracy are critical. The use of a CNN classifier ensures high detection accuracy while the selective processing of candidate images minimizes unnecessary computational overhead.
19. A non-transitory computer readable storage medium storing instructions, that when actuated by the processor, cause the processor to perform the method of claim 18 .
20. The method of claim 18 , wherein the processor comprises the first convolutional neural network (CNN), of plural CNNs.
Unknown
March 30, 2021
Browse 5M+ US patents with plain-English claim translations and AI-generated analysis.